Fundamentals of MIMO Wireless Communication

Prof. Suvra Sekhar Das

Department of Electronics and Communication Engineering
Indian Institute of Technology, Kharagpur

Lecture - 05
Wireless Channel Models – I

Welcome to the lecture on Fundamentals of MIMO Wireless Communication. In the previous

lecture we talked about layered view of transmitter and receiver. In that discussion we were
trying to say that if you take a look at the channel from the different points at a transmitter
you would see different manifestations of the link between the transmitter and the receiver.

(Refer Slide Time: 00:45)

If we take a look at the picture that we had drawn in the previous lecture, we had drawn the
schematic of the transmitter followed by the schismatic of the receiver while with the channel
in between. What we summarily said, is that typically we would look at the channel as the RF
channel for wireless communications. However, if we take a look at this section which
eliminates the RF section we are usually looking at analog baseband channels.

If we take a look at the previous section we are looking at the discrete samples of the channel.
That means, discrete samples go into the channels discrete samples come out of the channel.
There is a transmitter processing unit there is a receiver processing unit. the input to this are,
bits the output of these are again bits.
(Refer Slide Time: 01:41)

So, if we look at this part only what we have is the symbol mapper also sometimes called the
modulator which sends signal into a channel and channel gives output and it is the job of the
demapper to reconstruct the bits which were at the input of the symbol mapper. Here clearly
what we see inputs are constellation points, output is something in the signal space it can be
for example complex sequences could be any others. If we take any other stage, for example
if we are starting to look at the channel here that means, if I am taking sending signals in here
and getting them out the channel would appear as a binary channel.

What we are effectively trying to say is channel depends upon from where you are seeing it
and it encompasses the entire thing remaining on the right side of this particular picture. So
therefore, we can see that if we are talking about symbol mapping the job of the symbol
mapper and the demapper simultaneously is such, that it should generate signals in such a
fashion when it goes through this composite channel, is the composite channel of the RF
channel the up conversion, the down conversion, the D/A converter, the A/D converter. And
when it comes out after going through everything over here it should be able to reconstruct
that means, the demapper should be able to reconstruct whatever is happening here.

So, this channel depends upon what we have obstructed. The symbol mapping and the
demapper together is designed in such a way that there is minimal distortion happening to the
particular channel that it is observed. Symbol mapping is the module of the transmitter
demapper is the module of the receiver. This is usually a baseband signal processing unit and
this is usually a signal generation unit.

(Refer Slide Time: 04:13)

With this what we had also summarized is that when we talk about channels or the channels
which you have come across could be listed down as binary channel which you may have
come across as binary symmetric channels, binary erasure channels, instead of binary they
could be M-ary channels, it could be symmetric or asymmetric. These are usually probability
based and at symbol or bit level. Whereas, the next set of channels are usually the waveform
channel which takes continuous output. One such example of a famous channel is the
additive white Gaussian noise channel which you have come across in studies in digital
communications. Then there would be non-ideal channels as if there is a filter between the
transmitter and receiver, so here is the filter.

In the case of AWGN, there is addition of noise. In case of non-ideal there will be some kind
of a filter for example the telephone line. In non-ideal channels you would have time
invariant which is typically in wireline communications and time variant which is the most
important kind of channel in wireless communications. So, as we move ahead further in our
study of channels we will mostly be concentrating on time variant channels and usually these
models would be linear models.

Typically we say that the understanding of the channel is bread and butter for the wireless
communication engineer. The reason is clearly apparent because if we take a look at the
earlier picture especially this one the wireless communication engineer is supposed to design
individual modules at the transmitter he should be able to design individual modules at the
receiver. The job of the transmitting module is to generate signals which go through
minimum distortion. The job of the receiver module is to reconstruct the signal matching as
closely as possible to the input at the corresponding peer point.

So, unless we understand what is happening to the signal in this part of the channel with the
RF section or the analog section or the discrete analog section; we have to understand the
details then only we will be able to do a reconstruction. If we understand the details then we
can shape the signals in such a way that it undergoes minimal distortion. So, it is absolutely
essential to understand the communication systems. Typically before a communication
system starts getting designed the channel is usually measured and well understood, For
example, when 4G was getting designed or when 3G was getting designed one of the first
things was that happen to identify the frequency of operation.

Now again as you see over here, ultimately is the RF channel; ultimately it goes into the
wireless medium. This RF one of the fundamental elements which defines this is the carrier
frequency. And as we have said earlier if the frequency is let us say 700, 800 or 900 MHz this
is one common frequency which we know or if we look at 30 gigahertz, the effects of the
channel are different. Although we can remove the RF section in our analysis, the
corresponding effects need to be captured in the baseband channel. The effects which are at
900 MHz definitely different then the effects are 30 GHz.

A model we will be able to capture both provided appropriate parameters are used. And
therefore, all communication systems designed begins with first identifications of the
frequency band of operation so that the channel properties are well captured. Once the
frequency band of operation is decided then the channel which lies between the transmitter
and receiver is characterized.
(Refer Slide Time: 09:04)

So, basically channel characterization precedes any activity on the communication system.
So, when we mean channel characterizations what we mean is, understanding the channel
effects, what happens to the received signal when a particular time of signal is sent from the
transmitter. For example, we can study the effect on continuously of transmission or we can
also study the effects on impulse transmission, depending upon the kind of communication
systems that we are about to discuss. The other important thing we should note is when we
talk about the channel it depends upon what is our end goal. If we look at this part one
possible goal to design the communication system is to measure the bit error rate or to design
these blocks in such a way that the bit error rate is minimized. The channel model that will
produce should be able to provide sufficient details to capture the particular information that
we are supposed to be using in the design of communication systems.

Therefore what we can see is the channel is very very important; it is very important to know
what we are doing, it is very important to know what we are supposed to do, what exact
details we require from channel model. Now this channel modelling is a very very important
activity. We are generally not concerned with channel modelling most of the time in these
particular codes we will be concerned with channel models. Channel modelling is an activity
which precedes communication system design which requires channel sounding by sending a
particular training sequence measuring with very specific equipment coming up with curve
fitting and various methods by which a particular model can be generated.
Now let us take a look at what are the important aspects of studying the channel.

(Refer Slide Time: 11:24)

So, looking at the wireless channel models, understanding the channel model is important as
we have said it is very important for design of communication systems, where interested to
see what happens to the signal when it propagates through the particular channel. It helps us
design transmission signals as well as processing algorithms which we have just explained.
Also very importantly it provides links to fundamental limits on performance. What we mean
by this particular point is that when we are having a transmitter and a receiver and there is a
particular channel. Let us take one example where the transmitter and receiver is separated by
a distance of 10 meters the maximum radiated power that is available is let us say 0 dBm. So
we would like to understand, what is the maximum bit rate that can be achieved for this
particular system if the bandwidth is limited to a few megahertz?

If we take the same example and refine it further say our operating environment is the indoor
condition as case 1 and another the operating environment is outdoor, in a rich scattering
environment or let us say city centre region or if you take another region say the rural
scenario. In all these three cases although parameters are different the operating environment
are different and we would expect the performance limit to be different in each of the cases.

And if we add additional things for instant mobility if we say the transmitter and the receiver
are having relative motion and there is certain maximum speed between the transmitter and
receiver again that will influence the maximum bit rate that is achievable in this condition.
Very important thing about wireless communication systems is that the channel is random in
nature. We will see why it is and the variation of received signal or the received signal
strength is across the time domain. We have already said there is a time variant channel. The
received signal strength lies across frequency. For instance, I am interested in 900 MHz band
little bit to left of 900 MHz band and little bit to the right of 900 MHz the signal amplitude or
signal strength is going to be different.

Similarly, space is another important dimension and for us in this particular course space is
one of the most important dimensions that we will consider. Of course, time and frequency
are definitely taken into account. So, by space dimension what we mean is if I am receiving
the signal at a certain distance from the transmitter let us say 10 dBm or 10 dB of SNR and I
moved to a another location on the XY plane it could be moving closer to the transmitter or
far away from the transmitter or even the distance remaining the same we have changed the
angle, such as the radial distance between the transmitter and the receiver remain the same.

What we expect is that the average received signal strength could be different, simply
because the paths of signal which propagates from the transmitter to the receiver could be
undergoing different reflection refraction or scattering phenomena. There could be a line of
site, there could be a severely obstructed situation. These will decide the kinds of effects in
the signal undergoes.

Now, when we talk about the channel what we would be interested in is channel models. We
will not be generating the models, but we will not be using the exact channel properties, but
finally we will be using models. You have been using models in many studies. Now as the
name suggests it’s finally a modelling. A model means it is an abstraction of the exact
phenomenon that happens. For instance, again if we go back to this particular picture the
same channel is present over here. It depends upon what we are interested in according to the
channel is modelled differently.

Although it has the same effect it manifests itself at different locations in different ways. So
therefore, we should remember that we have end of the day using models and not the exact
channels.
(Refer Slide Time: 15:34)

Now why we need models, because many times the systems are very very complex and it is
not possible to get close form expression, it may not be possible to do even trials. For
example, if I am designing the communication systems I would like to build the transmitter I
have to build a receiver go and do field trials see what is the performance come back to the
lab design the transmitter again design the receiver again and do it. It is hugely time
consuming, it is very very complex. In order to avoid that we would like to use models so that
while sitting in a lab or sitting in a room or in an office we are able to design transmitter and
receiver components.

And many times one has to resort to simulations. We will see certain simulation things while
we are doing the course. By simulation means we imitate the effect that happens under
transmitter or at the receiver and what happens in the channel. So, once we send a signal that
goes out we model the impact of the channel and finally receive it. We can do all these things
using a digital computer or specialized equipment. So, we do not have to go to the real field
to the actual operation. We can do it right sitting in our labs.

I think it is the one of the most difficult part of studying wireless communications and this is
where we begin with. So, one of the messages which I would like to put forward is those who
all are aspiring to pursue this domain of wireless communication. It is very very important to
understand the propagation effects and the channel models. At least if you understand one
particular channel model you will get an idea how other channels are going to be, and due to
variation of the similar kind of models that you are going to encounter.

Models can be deterministic, they can be statistical. Deterministic we mean we solve EM

equations for every surface that the wave hits then it gets refracted reflected or diffracted or
even scattering and the account for the accumulating effect of these phenomena at the
receiver which could give rise to fluctuation of signal strength across time frequency and
space. These models are very very complex in nature, it requires a huge amount of signal
processing, and there are many many tools available for deterministic channel models; one of
them could be like ray tracing. However, in the methods which we use for a special base band
signal design we generally refer to statistical models.

These statistical models what they mean is there is a model with parameters and if we use the
models it adheres to the certain statistical measures, for example the mean, the variance, the
distribution or the coherence factors. So when the model, when iterated or tried over several
several iterations for example, millions of 10,000s of iterations the statistical properties
remains the same, and these are random in nature.

However, again the statistical models there could be analytical models there could be
empirical models. In empirical models what we have is channels are measured from
measurement curve fitting or in some other way models are developed which has the same
statistical parameter as that of the measured channel parameters. End of the day we need very
simple tools which we can use with pen and paper or on a computer so that we can design
communication systems.

In this course we will be encountering statistical channel models and mostly analytical. Of
course we will look at some of the empirical models, especially for large scale fading.
(Refer Slide Time: 19:16)

The factors which influence the channel are the propagation factors which are basically
reflection diffraction and scattering which we have been mentioning. We divide the study of
channel in the two broad categories; large scale fading and small scale fading. The large scale
fading are those effects or those models which capture the effect of a signal fluctuation over
large separation distances between the transmitter and the receiver, whereas small scale
fading are those models which capture the effect of signal propagation or signal strength
fluctuation across small separations.

If we consider this particular diagram as shown in this particular slide let us imagine that
there is a transmitter which sends out signals. When these EM waves propagate we will
assume of course the transmitter and the receiver are in the far field they are not in the near
field region. The signals propagate, they can fit a building surface they can come to the
receiver. There could be a direct line of site there could be reflection from the moving
objects, so it could be multitude of things that happen to the signal. So when they come to the
receivers they combine together give a combined effect at the receiver.

As a result of which there could be a signal strength fluctuations. So, when we studied small
scale fading we actually take into account this so called multi path propagation. The main
explains the propagations from the transmitter to the receiver happen through multiple paths.
That we first path, this could be this could be a second path, this could be third path and so
on. Theoretically there could be an infinite number of paths, but on all practical purposes we
generally model them finite number of paths. And that is usually dependent upon the
resolution of the instruments that we use to measure the channel characteristics.

The large scale fading is usually dealing with average signals strength, whereas the small
scale fading is usually to provide models which predict instantaneous signal strength
fluctuations. There is also the effect of mobility, because of the moving object between the
path of transmitter and the receiver which results in the Doppler shift we will see in detail.
And this causes time selectivity of the signal that means signals fluctuate in signal strength
even though the receiver may be static at one point. The receiver could be moving that could
be one case, the other case the receiver and the transmitter is fixed, but there are moving
objects which reflect refract or diffract the signals which finally come to the receiver.
Because of this there are time variations of the signals that are arriving at the receiver.

One thing we can grossly remember is that small scale fading phenomenon or this particular
model captures instantaneous variation of signals strength in the range of 30 to 40 dB
typically it is not the final numbered, but that is the usually the range of fluctuations of signal
strength at the receiver even though it is located at a fixed particular point.

(Refer Slide Time: 22:46)

When we study wireless communications or the channel models the wireless channel fading
effects that is a fluctuation of signal strength over time can be broadly categorized into small
scale fading and large scale fading as mentioned in a previous slide. The large scale fading
can be separated into two different models that is the path loss model and these shadowing
models which we will see in subsequent lectures.

The path loss models are also sometimes known as the attenuation model which describes the
loss of signal strengths due to separation of path length between the transmitter and receiver.
The shadowing on the other hand captures the effect of local fluctuations of average SNR and
improves upon the details of which is captured by the path loss model, both give the
prediction of average received signal strength.

If we look at small scale fading the small scale fading can be classified as time selective
fading, space selective fading and frequency selective fading. Time selective fading has the
name suggest there is selectivity that means selection and the selection across the time axis.

(Refer Slide Time: 24:09)

So, if we have time axis and if we have the received signal strength on this axis, across time
if the signal strength is very low the channel is not allowing signals to go through. Across
time if the signal strength is very high the channel would allow signals to go through
unattenuated. So, what we usually would see if I draw P rx, received power at a particular
distance as a function of time and measure it by let us say dBm, we might see this kind of
figures where it is like random fluctuations in time and these fluctuations would be in the
range of 30 to 40 dB.
We call it time selective because across time there are selections in time windows where the
signal passes through without much attenuation and there are sections in the time window
where there is huge amount of attenuation of the signal strength at the receiver. One typical
example that I can point at this particular instance is if you ever listen to amplitude
modulation radio let us say the medium wave or the short wave transmission if you tune into
a particular station you will often find fluctuating signal intensity, and that is one
representation of fading.

One thing I would like to remember again in this particular subject or the things that we deal
with it is very difficult to obtain situations where we can perceive the effects instantaneously.
So, it is a bit imaginative we have to extend our imaginations a little bit and that is why I
mentioned as simulations is one way of getting closer and moving beyond imaginations
towards practical systems. The next dimension is this space dimension. In the space
dimensions it can be rich scattering or poor scattering. In the frequency dimension it could be
frequency selective or flat fading. And the time dimension it could be fast fading and slow
fading.

We will discuss the details of fast fading and slow fading when we talk about small scale
fading. We will talk about details of frequency selective and flat fading when we talk about
more details about small scale fading. About space selectivity I will briefly tell you now and
we will take up in detail. We have rich scattering and poor scattering, environment a rich
scattering environment would be one where there is lots of reflections from the surroundings.

So, if we consider this particular room or the room where you are sitting the lot of walls, so
there is a generation of a signal from the transmitter there are reflectors all around and there
is a receiver let us say there could be one direct line of site path there could be reflected path
1, the reflected path 2, 3 could be multiple reflection and so on, so this is a rich scattering
environment. Whereas, if we have the transmitter and it is a rural region where there is a not
much obstruction available and there is a receiver and there could be one hill up will be a hill
some point, it could be mostly line of site may be one reflected path moving to it; so this
would be a poor scattering environment, this would be a rich scattering environment.

So, we will have to understand what goes on in these particular channel models so that we are
able to capture the effect of the channel.
(Refer Slide Time: 27:44)

Before we close today’s lecture I would like to explain this particular figure. This is available
in the book by Rappoport. In that figure it is shown that the separation between the
transmitter and receiver in the X axis represented in meters and the Y axis it is a received
power in dBm; what you see is the green lines fluctuating to a large extent is representation
of the instantaneous variation of the signal strength or it can be called the small scale fading,
because over small scale, this is a small scale within a meter which is within a few lambda or
a fraction of a lambda or the wave length is huge fluctuation of signal strength.

You can clearly see it goes down nearly -70 dB from around -37 dB or so. So, there is around
more than 30 dB of fluctuation in this particular snapshot, whereas if we start taking the
average of these instantaneous fluctuations in this range what we get is the red line. So, if we
are interested in the average or the model which predicts this average signals strength
fluctuations are captured under the large scale fading in models. Whereas, the small scale
fading model captures these small fluctuations, to show under large scale fading what we
have is the average fluctuation.

So, this in term predicts the average signal strength, this is talking about the instantaneous.
So, we will look at some details of large scale fading models which help us predict the
received signal strength as a function of transmitter receiver separation distance across large
distances. For instance, when the transmitter receiver separation distance increases by orders
of hundreds of meters or by orders of few kilometres, whereas when we will be interested in
studying the effect of signal strength fluctuation when the transmitter receiver fluctuation
separation is in the order of λ let us say a few centimetres or a few meters and then we will be
resorting to the small scale fading models.

In the next few lectures we will start the briefly taking a look at large scale fading models
before well being into details of small scale fading models. When we are concerned with the
transmitter and receiver design we would be mainly interested in the small scale fading. For
the sake of completeness we will briefly take a look at large scale fading models and just let
me remind you our objective is to look at large scale fading models for the sake of
completeness, but what we will be mostly needing in this particular course is the small scale
fading models.

The reference books that you can follow for the future few lectures will be Wireless
Communications by Rappoport, Principles of Mobile Communication by Gordon Stuber,
Wireless Communications by Andrea Goldsmith and many others. So, I would again request
you to keep the references handy. Of course we will provide you with sufficient details in this
and we may be able to provide you with some of the slides, but always reading such
references would refine and provide more clarity to the details, because again iterate this is
one of the critical parts of communications.

And successes in this course would depend primarily on understanding the details of these
models or what is the abstraction that has been done in this model because the rest of the
details MIMO communications would be using these models which we will not refer to a any
further details once we cross over this section on this propagation effects of the wireless
communication links, thanks.

Thank you.